fp.

Hi HN, we're Aarush and Arnav from The Subvocal Company (YC F26). We’re building smart glasses that let you talk to your computer without talking out loud. You can mouth words silently and discreetly and the glasses decode them to text at the speed of speech.

Here's a video of Dave from YC trying it out: https://www.youtube.com/watch?v=9QdX1eObVYI

(Edit: you don't actually have to move your lips or even open your mouth for this to work! Since the RF can see inside your mouth the system can be fine tuned on "internal speech". This isn't the automatic tongue movements that occur when "thinking" words to yourself, but more like mouthing with your lips closed. It's a bit weird to do, but with practice, we have been able to get it down pretty well.)

Arnav and I were studying CS at Georgia Tech at the beginning of the year, with no experience in silent speech or sensing. We were both AI-native builders and power users of Wispr Flow. What bugged us was not being able to use these tools in the library (or an office) or in public. At Stanford’s TreeHacks in February, we decided to build the most technically ridiculous thing we could think of. We’d read about AlterEgo, a silent speech system out of MIT’s Media Lab, and wanted to recreate it in 36 hours. We built a crude prototype, sticking EMG electrodes onto my face:

https://drive.google.com/file/d/1mwhv5HZMz3UTevdrgwS5O7g2jWV...

We got it to kind of work at a letter level in a very janky way. It wasn’t good enough to win the hackathon, but convinced us to go all in. The first time I typed a sentence without anyone noticing, I felt the magic of the technology.

We quickly realized EMG (reading muscle electrical signals) wouldn’t work:

- Signal changes due to sweat

- Electrical interference

- Placement variability

- Most importantly, no one wants to stick electrodes onto their face!!

So we switched to RF and built the glasses you see now. They emit radio waves, similar to Bluetooth, that pass through your face. As they cross the tongue, lips and soft palate, they’re modulated by silent mouthing. We train foundation models on this data to decode these perturbations to text.

Despite prior academic research, getting this to work was a huge engineering challenge. First we built an extremely low noise RF sensing stack. Then we had to place antennas so their fields overlapped maximally with the articulators. Our first attempt on glasses failed completely, so we reverted to old habits and taped antennas on our face:

https://drive.google.com/file/d/1ks0dS0eZiTV9EqSFsyB_w5BBjpc...

Since taping antennas consistently in the same spot and angle was hard, we created this weird but effective head rig to hold them in place:

https://drive.google.com/file/d/1rhzNFpdDzr7oHk0isFY70dkoV4C...

This worked, but obviously wasn’t a product yet. So we worked deep in simulations to design glasses that could hold the antennas.

That evolved into our design now, with antennas embedded within the frames. Once we established there was a signal, the next hurdle was cross-user generalization. Everyone’s facial anatomy, and hence RF signal patterns, are vastly different. We tried algorithmic workarounds but realized the only lever was to massively scale pretraining data.

So that’s what we did. We set up booths in our office and paid people to read prompts wearing our glasses, to pretrain our models on. This produced a beautiful scaling curve, with errors falling quickly as we scaled our dataset and cross-user performance becoming genuinely compelling (as you see in the video above; Dave only had a short 20-sentence calibration).

We always heard about building things from first principles but never understood it before this. We tried tackling it top down but completely failed. Only when we broke each component (the antenna, electronics, ML) down to its fundamentals to understand and optimize it did we get it to work really well.

Over the next few months we’re continuing R&D to turn our prototype glasses into the slick product on our website and ship in Q1 2027. We’re also scaling up our dataset to improve performance on edge cases.

For anyone who loves voice dictation but hates using it around people, we’ve opened preorders for our first batch of 500 units, with a $200 discount off the final price. Allocation is running out quickly, so get yours at https://subvocal.co/preorder!

Comments

hmokiguess•50m ago

EDIT: I find talking in that manner in a public setup to be awkward and unpleasant, maybe that's just me.

That said, I think the idea of having a private walkie-talkie that connects on its own frequency for large crowded spaces to be very cool

dang•50m ago

> I like it but I don't want it, I'm not sure, it feels weird and cringy.

"Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something."

https://news.ycombinator.com/newsguidelines.html

p.s. Especially please don't do this when a thread is brand new. Such comments are the easiest to post, and therefore the quickest to appear, but by posting them, you steer the discussion straight away from the curious, thoughtful conversation we want here. Threads are extremely sensitive to initial conditions, so this is important.

hmokiguess•49m ago

Apologies, I edited, can you tell me if my edit is ok? EDIT: happy to delete and repost my edit instead too

dang•46m ago

It's better. I've turned off the flag.

theturtletalks•50m ago

Someone built an open-source version of this, but it’s far from polished:

https://github.com/joeblau/subvox

There was this really interesting one where it reads your lips:

https://github.com/amywork777/lipflow

I’m using these to build my own so I can work in co-working spaces, planes, and libraries. Though moving my mouth with no words might draw some eyes.

willio58•46m ago

> There was this really interesting one where it reads your lips

This is what I thought the original post was about. The glasses you have to wear is where that lost me.

If we could get lip-reading via your computer/phone camera down perfectly, that's the real win. Making people wear your goofy frames is where you lose a massive portion of people. Nobody wants to look like Zuckerberg lol

aarushyadav•40m ago

Lip reading is one of the other modalities we tried as well beyond IMU, EMG and RF! We worked a lot on fine tuning Auto-AVSR, one of the most commonly used open source lipreading models. It's very enticing to work on, since it gets to about 17% word error rate, which looks good enough that you would think working on it further would get it down to product quality, but we found there's an information-theoretic limit on the modality. Speech has 40 phonemes, but only about 12 to 15 visemes (externally visually separable phonetic states). For example P and B look very similar from a camera. This causes the scaling laws on lipreading to flatten out at an error that's too high to be product ready but just low enough to make for a good demo. Our original plan was actually to ship a lipreading software to see if people would want this behavior, but we realized you couldn't make lipreading work!

zie•25m ago

Any deaf/HoH person who has tried to do lipreading could have told you all of this too.

Good luck on the project!

theturtletalks•14m ago

Now that is insightful! The demo definitely fooled me and even the first repo requires you to train the model.

Lip reading AI would be pretty big since privacy would probably go out the window.

aarushyadav•43m ago

Yup we worked with IMUs as well when we started this. They measure aggregated face motion, so cannot distinguish between fine phonetic classes, only broad words. The open source demo got to a 5 word classifier, I believe. We tried a lot to make IMUs work too, in various form factors, but the signal doesn't contain enough information for open ended decoding beyond classifying between a few words. We would have loved to be able to just push software!

OhMeadhbh•43m ago

Spent 30 seconds scrolling down to try to figure out what this is. Eventually clicked away from the site because it was taking too long to get to the point. I'm sure it's great and I understand it's something to do with glasses and subvocalizations, but consider a one sentence contextualization at the top of the page.

LatencyKills•26m ago

I was just trying to help a vibe coder and immediately noticed that their README.md was over 2,000 lines long. I asked, "Why is your readme so long?" and their response was, "This is just a really complicated application."

The app essentially manages a bunch of TODO items written in markdown.

derefr•16m ago

For some reason AI coding agents seem to think a README should be a Product Requirements Document (PRD).

I think it's because most projects don't have any settled place (or at least, any place that lives inside the project's source-code repo) to record PRD-like information.

But because AI coding agents really do benefit from having PRD-like info (i.e. a hierarchical breakdown of the "intended what" and "intended how" of the system that "should" exist in the repo, kept updated to reflect changing ideas about those intentions), they really want to record that type of info somewhere, so they can then reference it.

AI coding agents don't seem to have any "taste" (or maybe "confidence"?) when it comes to asserting that even if you don't already have a separate PRD-like file in the repo, you should have one, if you want the agent to do its job well (and so it's just going to create one.) Instead, they just shrug and cram the PRD stuff into whatever structure you do already have. Which, if you're starting fresh with a greenfield project, is usually just a README.

(I hypothesize that if you tell an AI coding agent to use Behavior-Driven Development [BDD] on a greenfield project, it'll probably lean toward maintaining its PRD-like info in the form of BDD test cases. Haven't tried this experiment yet; would be interested to know the outcome!)

LatencyKills•8m ago

That is exactly what happened in this case; the readme was the PRD.

Throughout the document Claude kept claiming that issues were "mitigated but not fixed". They were essentially bandaid fixes for things it couldn't figure out. When I pointed this out to the author he said, "Well Claude thinks it works."

I wonder what happens to software when enough people try to ship projects like this.

aarushyadav•24m ago

Yup that makes sense. Thanks for the feedback, we'll make some changes shortly!

jasonlotito•20m ago

> Talk without talking. > Input text at the speed of speech, silently and invisibly.

I understood from just this. But the demo video impressed even further.

> Spent 30 seconds scrolling down to try to figure out what this is.

Did you try reading the site? Or maybe doing something other than just scrolling?

Does it need some form of subway surfer to get you to pay attention?

> Eventually clicked away from the site because it was taking too long to get to the point.

So yeah... subway surfer. This generation is cooked.

avree•26m ago

Do most people talk faster than they can type? It's the opposite for me - I also read aloud slower than I can read in my head.

aarushyadav•20m ago

Most people type at 40-50 WPM. Developers can often type at 70-100 WPM. Meanwhile, most people can talk at 120-200 WPM. Beyond just speed, we've found with many people that talking things out tends to flow more naturally, since it's often a lower friction action so people don't "sanitize" the text in their head before outputting it. Leads to better prompts quite often :)

achandra03•10m ago

> talking things out tends to flow more naturally,

I've found the opposite. Typing feels much more "deliberate", and of course you can hit backspace but you can't un-speak a word. With prompting, I've found that, in general, speaking leads to lower-quality results than typing.

cogythea•8m ago

The bit before your quote was

> with many people

Not everyone. I, at least, definitely overthink what I type, and while I end up with nice-ish sentences in my prompts it takes a lot longer than a train of thought that the AI parses correctly most of the time anyway.

xbar•19m ago

Most people talk faster than they can type. 80 wpm voice chat is easy for everyone. 80 wpm typing filters out most typists.

mrngld•2m ago

This has been an interesting thing for me to watch as it's such a case of people having widely different experiences. I'm sure people that didn't grow up on a keyboard like me can probably talk quicker than they type, but even aside from me being a fairly good typist my speech is held up by being a bit more halting, a bit more of backtracking on word choice, sometimes not quite the same 'voice' of writing as I'm aiming for, all of which obliterates any speed advantage it'd have for me.

But happy to see people get more options.

This reminds me of Eternity's End by Jeffrey Carver and the implants they had. Released in 2000 but the more time goes on the more the neural implant technology is coming about. To be fair I think Enders Game had similar tech to detect subvocal speech like this product.

lstodd•1m ago

Writing also filters the useless crap that is talked but is not worth writing down.

Zambyte•16m ago

I'm not reading aloud most of the time that I'm talking.

egeozcan•16m ago

Yes, I'd estimate the vast majority of people speak faster they can type. Not sure about the thinking part because I usually have 20 threads going parallel in my head and can't imagine what normal thinking would be like.

sspiff•13m ago

I'm the same. I also can think ahead / along much better while typing. When I'm talking I'm just talking, not thinking as much.

But I guess thinking is going out of style anyway.

That said, I know a lot of people prefer talking to their computer, especially when their computer is a phone. So there is definitely a place for stuff like this.

rickydroll•11m ago

You need to train yourself to speak written speech. When you do that, the thought processes behind typing your words activate, and your language use changes.

jbs789•4m ago

How?

lstodd•3m ago

Why bother when old ways work _and_ there is no benefit? Or name one.

wongarsu•7m ago

While sitting at the desk, I type about as fast as I talk. While unloading the dishwasher I'm a much faster talker than typer